Comparing Early Graft Dysfunction Scores in Liver Transplantation After Ex‐Situ Normothermic Machine Perfusion Authors: Kejal Shah, Sydney Castellanos, April J Logan, Leonid Gorelik, Annelise Nolan, Manoj Iyer, Lauren Von Stein, Wei Chen, Ayato Obana, Ashley Limkemann, Navdeep Singh, Sylvester Black, William K Washburn, Austin D Schenk, Musab Alebrahim, Sai Rithin Punjala Published: 2026 ABSTRACT ABSTRACTIntroductionAs normothermic machine perfusion (NMP) maintains liver grafts in a near‐physiologic state, uncoupling ischemia–reperfusion injury from the recipient, the validity of early graft dysfunction scores from the static cold storage (SCS) era are uncertain.MethodsAll liver grafts that were transplanted following NMP at our center between 08/19/2022 and 05/31/2024 were analyzed. We assessed associations between the early graft dysfunction scores and graft survival in the NMP era and identified factors linked to pronounced dysfunction.ResultsOf 126 NMP‐perfused liver grafts, seven (5.5%) were discarded after viability assessment. Among the 107 transplanted livers (41%‐DBD, 59%‐DCD), neither the liver graft assessment following transplantation (L‐GrAFT7) nor the model for early allograft function (MEAF) scores correlated with early graft loss. Early allograft dysfunction (EAD) occurred in 40% of cases and was associated with lower 6‑month graft survival compared with non‐EAD cases (89% vs 100%, p = 0.04). Higher donor BMI, longer agonal and anastomosis times, post‐reperfusion syndrome, and suboptimal pump parameters (impaired glucose metabolism, delayed lactate clearance, increased bicarbonate use, elevated perfusate transaminases) correlated with EAD. Grafts exceeding conventional viability thresholds (increased bicarbonate use to maintain pH >7.2, lactate rise >2 mmol/L within 6 h post‐clearance, and perfusate ALT >6000 U/L) were successfully utilized without compromising short‐term outcomes.ConclusionIn the NMP era, EAD identifies a high‐risk subgroup with inferior short‐term graft survival. Trajectory‐based scores did not correlate with early graft loss. Pump variables may help guide decision‐making when considering the use of liver grafts with an increased likelihood of EAD in select recipients.Keywords: allograft dysfunction, liver transplantation, machine perfusion, viability Abbreviations aspartate transaminase alanine transaminase donation after brain death donation after circulatory death donor risk index early allograft dysfunction endoscopic retrograde cholangiopancreatography intensive care unit metabolic dysfunction‐ associated steatohepatitis model for end‐stage liver disease‐ sodium ex‐situ normothermic machine perfusion in‐situ normothermic regional perfusion primary biliary cholangitis post‐reperfusion syndrome primary sclerosing cholangitis rejection activity index static cold storage Introduction Deceased donor liver transplant volumes have risen sharply over the last few years in the United States, particularly from donation after circulatory death (DCD) donors [1]. Normothermic regional perfusion (NRP) and ex‐situ normothermic machine perfusion (NMP) have played a dominant role in expanding the donor pool, improving outcomes, enabling better resource utilization, and more efficient transplant logistics, leading to wider adoption of these technologies among centers [2, 3, 4, 5, 6]. With the use of perfusion technologies, the decision to use an organ is not solely based on conventional donor and recipient characteristics, and the ischemic interval incurred during cold preservation. Currently, the decision to use a liver graft includes real‐time organ‐specific quality data [7, 8]. In the era of static cold storage (SCS), scoring systems including the liver graft assessment following transplantation (L‐GrAFT) score, the model for early allograft function (MEAF) score, and the criteria for early allograft dysfunction (EAD) were validated to risk‐stratify recipients for poor posttransplant outcomes [9, 10, 11]. By maintaining the liver graft in a near‐physiologic state, NMP fundamentally alters organ preservation, mitigating severe ischemia reperfusion injury (IRI), reducing rates of primary non function, and improving graft and patient outcomes [2, 12, 13, 14]. This challenges the current utility of dysfunction scores that were developed in the cold storage era. The primary aim of this study was to evaluate the association between early graft dysfunction scores (EAD, MEAF, and L‐GrAFT7) and the key clinical outcomes in the NMP era. The secondary aim was to evaluate the correlation between NMP perfusion parameters and graft dysfunction scores with predictive value, in the NMP era. Patients and Methods Study Population and Data Source We performed a retrospective cohort study of all deceased donor liver transplants performed at our center between August 20, 2022 and May 31, 2024. Electronic medical records were reviewed to collect the data. All DBD and DCD liver grafts that underwent ex‐situ NMP were included in the study. All liver grafts were transported back to the recipient center using SCS and then placed on the device and perfusion was initiated at the recipient hospital (i.e., back‐to‐base NMP). Pediatric transplants, recipients of multi‐organ transplants, living donor liver transplants, DCD liver grafts that underwent NRP or NRP followed by ex‐situ NMP (four liver grafts), and deceased donor liver transplants performed with SCS were excluded. Data on bacterial and fungal infections in the recipients, acute cellular rejection, and biliary complications were obtained from a manual chart review of the first 6 months of postoperative data (through November 25, 2024). Graft and patient survival were based on follow‐up through September 12, 2025 (Kaplan–Meier median follow‐up days 666, 95% confidence interval 615–701 days). Study Groups and Definitions MEAF and L‐GrAFT7 scores were calculated using established formulas, and EAD was defined from their respective validation studies [9, 11, 15]. EAD was defined as the presence of one or more of the following within the first 7 postoperative days‐ bilirubin ≥ 10 mg/dL on postoperative day 7, INR ≥ 1.6 on postoperative day 7, or ALT or AST > 2000 U/L within the first 7 days. MEAF score is a continuous score (range 0–10) derived from the maximum ALT, INR, and bilirubin during the first 3 postoperative days, each transformed by a published logistic formula and summed, with higher scores indicating worse early function. L‐GrAFT7 is a score calculated from the trajectory (area under the curve) of AST, bilirubin, INR, and platelet count over the first 7 postoperative days, modeling cumulative early allograft injury and synthetic dysfunction. The quality of donor liver grafts was defined using the donor risk index (DRI) for the liver transplantation calculator [16]. Agonal time or functional warm ischemia time (fWIT) was defined as the duration of the donor systolic blood pressure (SBP) <80 mmHg. Agonal time for DCD acceptance was ≤ 30 min. Post‐reperfusion syndrome was defined as a >30% drop in the mean arterial pressure for >1 min within 5 min of reperfusion. Post‐reperfusion wedge liver biopsies were used to assess donor liver quality, as donor liver biopsies were not routinely performed at donation [17, 18]. Anastomosis time was defined as the time between “liver out of ice” after the end of NMP and reperfusion with portal blood. Cold ischemia time (CIT) was defined as the time from the cross clamp to the start of perfusion on the normothermic machine. Ischemic cholangiopathy was identified and classified as follows: grade A, normal; grade B, diffuse necrosis; grade C, multifocal progressive; grade D, confluence‐dominant; and grade E, minor form [19]. Liver grafts on NMP were discarded if there were concerns regarding macroscopic appearance, had poor flows (<150 mL/min arterial flows and <500 mL/min portal flows) or failed to clear lactate to <2 mmol/L within 6 h of perfusion. Blood gas data were available for all liver grafts that underwent NMP. Supplemental bicarbonate was administered to maintain pH >7.2 throughout perfusion. Increased bicarbonate needs to maintain a pH >7.2 was considered a relative contraindication if lactate metabolism was slow. The importance of bicarbonate needs was not appreciated initially; therefore, data on bicarbonate provision were available for 89.7% (n = 96) of NMP‐perfused livers in the study. Similarly, perfusate transaminase level measurement began after the first 20 liver grafts that underwent NMP. Perfusate transaminases levels were measured at the end of perfusion and were considered a relative contraindication to liver use when there were other ongoing concerns about viability. Bile production and chemistries were not included in the viability assessment during the study period. Statistical Analyses To examine downstream outcomes associated with posttransplant injury scores, recipients were grouped into “high” and “low” categories based on the median L GrAFT7 and MEAF scores (Figure S1). Cohort size prohibited quartile‐based stratification. One recipient died on postoperative day four and was excluded from the L‐GrAFT7 calculation. Kaplan–Meier curves were generated for graft and patient survival and were stratified by EAD status, L GrAFT7, and MEAF groups. Survival time was calculated from transplant to graft failure or death, censored at the last date of chart review (September 12, 2025). Death censoring was unnecessary, because no deaths occurred without preceding graft failure. Differences between groups were tested using the log rank test. Pointwise comparisons of survival probabilities at 3 months, 6 months, and 1 year were evaluated using Z tests based on the standard errors from the Kaplan–Meier estimators; only statistically significant comparisons were reported. As only EAD status was associated with significantly lower graft survival at 6 months, further subgroup analyses were limited to identifying the baseline donor, recipient and pump variables associated with EAD. Counts, percentages, medians, and interquartile ranges (IQRs) were summarized and stratified according to EAD status. Group differences were evaluated using the Mann–Whitney U test for continuous variables and Fisher's exact test for categorical variables because of non‐normal distributions and small sample sizes. Perfusion data were collected beginning at a 15‐min baseline point and subsequently at hourly intervals for up to 17 h. To identify the perfusion thresholds most strongly associated with EAD, we evaluated the rate of change in lactate and glucose levels over time versus baseline values. Cut‐points of −80% for lactate at 1 h and −10% for glucose at 3 h were selected by maximizing the discrimination between EAD and non‐EAD recipients using the area under the receiver operating characteristic curve (AUC) across hours and delta values (Table S1). To assess the relationship between perfusion transaminases and early posttransplant injury, regression lines were fitted between pump ALT and peak ALT within 7 days posttransplant, and similarly for AST. Both pump and posttransplant transaminases were log10‐transformed prior to analysis. Predicted values and 95% confidence intervals from the fitted models were visualized alongside scatter plots to distinguish between EAD and non‐EAD recipients. Because of the limited sample size and variability in available data across time points, pairwise deletion was used for missing data in the descriptive statistics and univariate tests. All analyses were performed using SAS version 9.4 (TS1M8; Cary, NC, USA). Statistical significance was defined as p <0.05. IRB approval was obtained from The Ohio State University (Protocol #2023H0246). The requirement for individual informed consent was waived owing to the retrospective nature of the study. All research was conducted in accordance with both the Declarations of Helsinki and Istanbul. Results Discarded Liver Grafts A total of 126 liver grafts underwent NMP during this study period. A total of 11 (8.7%) liver grafts (8 DCD, 3 DBD) were discarded after undergoing NMP, of which seven liver grafts (5%) were discarded based on viability assessment alone (Figure S2). One liver graft each (both DCD) was discarded because of poor arterial flows and uncontrolled surgical bleeding on the pump from recovery injury. Two liver grafts (both DCD) were discarded because of unacceptable biopsy results only (fibrosis), which resulted after the livers were placed on the pump. Four liver grafts (three DCD, one DBD) were discarded as they failed to clear lactate to <2 mmol/L within 6 h of perfusion/rising lactate. Two liver grafts (1 DCD, 1 DBD) were discarded due to rise in lactate to >2 mmol/L within 6 h, after initial lactate clearance to <2 mmol/L. One liver graft (DBD) was discarded due to a very low perfusate glucose level (1 mg/dL). Characteristics, Incidence, and Outcomes of Early Graft Dysfunction Scores A total of 115 liver transplants were performed after NMP during this study period. After excluding four multi‐organ transplants and four liver grafts that underwent NRP followed by NMP, 107 liver transplants were analyzed (44 [41%] DBD and 63 [59%] DCD). EAD occurred in 43 (40%) patients. The median LGrAFT7 score was ‐3.25, and the median MEAF score was 5.95; scores above these median values were categorized as high L‑GrAFT7 and high MEAF, respectively. There was no difference in the overall patient survival between the EAD and no EAD groups (log‐rank p = 0.15), between the high and low LGrAFT7 groups (log‐rank p = 0.68), or between the high and low MEAF groups (log‐rank p = 0.50). Similarly, there was no difference in the overall graft survival between the EAD and no‑EAD groups (p = 0.29), high vs low L‑GrAFT7 (p > 0.99), or high vs low MEAF groups(p = 0.79). However, 6‑month overall graft survival in the EAD group was significantly worse than in the non‑EAD group (89% vs. 100%, p = 0.036) (Figure 1). Patient and graft survival outcomes of early graft dysfunction scores. Kaplan–Meier plots depicting patient and graft survival showed significant difference between the EAD and no EAD groups at 6 months (p‐value = 0.04). However, the overall graft and patient survival were not significantly different between the groups. Baseline Characteristics Associated With EAD When comparing donor variables between the EAD and non‑EAD groups, only donor BMI (31.12 vs. 26.23 kg/m2, p < 0.001) and agonal time in DCD livers (19 vs. 15 min, p = 0.037) were significantly different between the groups. There were no significant differences in DBD/DCD status, donor age, sex, HCV status, terminal AST/ALT, cause of death, CIT, NMP duration, total preservation time, or degree of macrovesicular steatosis, fibrosis, or necrosis on reperfusion biopsy (Table 1). TABLE 1Baseline characteristics associated with EAD.Total (N = 107)EAD (N = 43)No EAD (N = 64)NMedian (IQR) or n (%)NMedian (IQR) or n (%)NMedian (IQR) or n (%) p‐valueDonor characteristicsDonor typeDBD10744 (41.1)4316 (37.2)6428 (43.8)0.5516DCD63 (58.9)27 (62.8)36 (56.3)Age, years10749 (36–58)4351 (42–60)6448 (35–58)0.1788SexMale10767 (62.6)4330 (69.8)6437 (57.8)0.2286Female40 (37.4)13 (30.2)27 (42.2)BMI, kg/m210728.09 (22.91–32.99)4331.12 (25.29–37.59)6426.23 (22.10–29.74) 0.0004 US DRI1072.08 (1.62–2.56)432.12 (1.72–2.60)642.01 (1.59–2.48)0.3278DBD US DRI441.47 (1.21–1.88)161.47 (1.34–1.96)281.49 (1.17–1.85)0.4067DCD US DRI632.43 (2.08–2.74)272.43 (2.11–2.65)362.43 (2.06–2.74)0.8731Donor HCV statusNAT(+)1076 (5.6)431 (2.3)645 (7.8)0.3980NAT(‐)101 (94.4)42 (97.7)59 (92.2)Last AST pre‐donation, U/L10752 (27–95)4355 (30–108)6447 (26–92)0.6224Last ALT pre‐donation, U/L10740 (22–86)4354 (24–79)6439 (20–91)0.6564Peak total bilirubin pre‐donation, mg/dL1070.5 (0.4–0.8)430.5 (0.4–0.8)640.6 (0.4–0.8)0.3016Cause of Donor DeathAnoxia10764 (59.8)4328 (65.1)6436 (56.3)0.1055CVA/Stroke23 (21.5)9 (20.9)14 (21.9)Head/Trauma14 (13.1)2 (4.7)12 (18.8)Other6 (5.6)4 (9.3)2 (3.1)Agonal time, minutes (DCD only)6216 (12–21)2719 (13–24)3515 (11–19) 0.0371 NMP CIT (hr)1075 (4–5)435 (4–6)645 (4–5)0.6542NMP duration, hours10710 (8–12)4311 (8–14)6410 (8–12)0.2409Total preservation time, hours10715 (13–17)4315 (13–18)6415 (13–17)0.1519Macro steatosis, yes7661 (80.3)3125 (80.6)4536 (80.0)1.0000<5%6136 (59.0)2512 (48.0)3624 (66.7)0.34985%–33%21 (34.4)11 (44.0)10 (27.8)34%–66%4 (6.6)2 (8.0)2 (5.6)Fibrosis, yes9215 (16.3)385 (13.2)5410 (18.5)0.5755Stage 11513 (86.7)54 (80.0)109 (90.0)1.0000Stage 22 (13.3)1 (20.0)1 (10.0)Necrosis on reperfusion biopsy9089 (98.9)3636 (100.0)5453 (98.1)1.0000<5%8918 (20.2)367 (19.4)5311 (20.8)0.58375%–33%53 (59.6)20 (55.6)33 (62.3)34%–67%14 (15.7)6 (16.7)8 (15.1)>67%4 (4.5)3 (8.3)1 (1.9)Positive Donor Blood Cultures1069 (8.5)434 (9.3)635 (7.9)1.0000Recipient characteristicsAge, years10756 (47–63)4360 (51–65)6455 (46–61)0.0793SexMale10767 (62.6)4325 (58.1)6442 (65.6)0.5414Female40 (37.4)18 (41.9)22 (34.4)BMI, kg/m210729.12 (25.43–32.25)4329.46 (25.73–32.46)6427.78 (25.12–32.11)0.2072MELD Na10720 (16–24)4318 (15–22)6421 (18–26) 0.0074 Cause of liver diseaseAlcohol10750 (46.7)4318 (41.9)6432 (50.0)0.3209NASH30 (28.0)15 (34.9)15 (23.4)PSC/PBC/ autoimmune10 (9.3)2 (4.7)8 (12.5)Other17 (15.9)8 (18.6)9 (14.1)Hepatocellular carcinoma10730 (28.0)4313 (30.2)6417 (26.6)0.8265Redo liver transplant1074 (3.7)431 (2.3)643 (4.7)0.6472Baseline creatinine, mg/dl1071.06 (0.81–1.40)431.00 (0.80–1.28)641.09 (0.82–1.42)0.4885Baseline eGFR ml/min/1.73m2 10775 (52–90)4381 (51–90)6474 (54–90)0.6189Pre‐op dialysis, yes1075 (4.7)432 (4.7)643 (4.7)1.0000Wait‐time to transplant, days10739 (8–175)4372 (9–202)6429 (8–139)0.3752Operative characteristicsOperation time, minutes107305 (259–371)43301 (267–371)64308 (250–371)0.9747Start Time of surgery8pm‐ 4 am1076 (5.6)435 (11.6)641 (1.6)0.07004am‐ 12pm83 (77.6)30 (69.8)53 (82.8)12pm‐ 8 pm18 (16.8)8 (18.6)10 (15.6)Anastomosis time, minutes10727 (24–32)4329 (26–33)6427 (23–32) 0.0429 Intra‐op RBC transfusion, units1076 (3–9)436 (3–9)646 (3–9)1Intra‐op platelet transfusion, units1072 (1–3)433 (2–4)642 (1–3)0.1664Intra‐op FFP transfusion, units1076 (3–9)436 (4–8)646 (3–10)0.9771Intra‐op Cryoprecipitate transfusion, units1073 (2–5)433 (2–4)644 (2–6)0.3357Post‐reperfusion syndrome10725 (23.4)4317 (39.5)648 (12.5) 0.0021 Post‐reperfusion cardiac arrest1071 (0.9)431 (2.3)640 (NA)0.4019Abbreviations: BMI, body mass index; CIT, cold ischemia time; CVA, cerebrovascular accident; DBD, donation after brain death; DCD, donation after circulatory death; DRI, donor risk index; EAD, early allograft dysfunction; eGFR, estimated glomerular filtration rate; FFP, fresh frozen plasma; MASH, metabolic dysfunction‐ associated steatohepatitis; MELD‐Na, model for end‐stage liver disease‐ sodium; PBC, primary biliary cholangitis; PSC, primary sclerosing cholangitis; RBC, red blood cell.Bold values show statistically significant. Baseline characteristics associated with EAD. Abbreviations: BMI, body mass index; CIT, cold ischemia time; CVA, cerebrovascular accident; DBD, donation after brain death; DCD, donation after circulatory death; DRI, donor risk index; EAD, early allograft dysfunction; eGFR, estimated glomerular filtration rate; FFP, fresh frozen plasma; MASH, metabolic dysfunction‐ associated steatohepatitis; MELD‐Na, model for end‐stage liver disease‐ sodium; PBC, primary biliary cholangitis; PSC, primary sclerosing cholangitis; RBC, red blood cell. Bold values show statistically significant. Among recipient and intraoperative variables, MELD‐Na score was lower in the EAD group (18 vs. 21, p = 0.007), anastomosis time was longer (29 vs. 27 min, p = 0.04), and post‑reperfusion syndrome occurred more frequently (39.5% vs. 12.5%, p = 0.002). There were no differences between the EAD and non‑EAD groups in recipient age, sex, BMI, primary liver disease etiology, history of hepatocellular carcinoma, re‑do liver transplantation, baseline renal function, wait‑time to liver transplant, total operative duration, proportion of daytime surgeries, or intraoperative blood transfusion requirements (Table 1). Pump Variables Associated With EAD The median duration of NMP was 10 h (IQR, 8‐12 h). The median perfusate pH, lactate, and glucose levels over time for all liver grafts that underwent NMP, stratified by EAD status, are presented, highlighting cases that resulted in graft failure within 6 months. (Figure S3). Multiple NMP parameters were significantly associated with EAD. The amount of sodium bicarbonate administered (45 vs. 30 mEq, p = 0.005), rate of sodium bicarbonate use (total sodium bicarbonate by NMP duration) (4 vs 3 mEq/h, p = 0.02), perfusate AST (4586 vs. 1771 U/L, p < 0.001) and perfusate ALT level (2518 vs. 897 U/L, p < 0.001) were all higher in the EAD group. Additionally, impaired glucose metabolism (<10% decrease in perfusate glucose concentration within 3 h of perfusion) (41.5% vs. 9%, p < 0.001) and impaired lactate clearance (<80% decrease in perfusate lactate concentration within one hour of perfusion) (78% vs. 58%, p = 0.04) were more common in the EAD group. There was no difference in the proportion of liver grafts with lactate rebound (rise to >2 mmol/L within 6 h after initial clearance to <2 mmol/L) between the EAD and non‐EAD groups (14% vs 20%, p = 0.449) (Table 2). Table S1 shows the AUC by hour, percentage change, and maximum AUC for each variable. TABLE 2Pump variables associated with EAD.Total (N = 107)EAD (N = 43)No EAD (N = 64) p‐value N Median (IQR) or n (%) N Median (IQR) or n (%) N Median (IQR) or n (%)NMP pump variablesTotal bicarbonate given, mEq9640 (23–63)3945 (35–75)5730 (20–50) 0.0048 Bicarbonate >40 mEq9641 (42.7)3921 (53.8)5720 (35.1)0.0929Total bicarbonate by NMP duration (mEq/h)964 (2–6)394 (3–7)573 (2–5) 0.0180 Lactate rebound (rise in lactate >2 within 6 h after initial clearance)10719 (17.8)436 (14.0)6413 (20.3)0.4498Less than 80% decrease in lactate by hour 1.10569 (65.7)4132 (78.0)6437 (57.8) 0.0373 Less than 10% decrease in glucose concentration by hour 310523 (21.9)4117 (41.5)646 (9.4) 0.0002 Perfusate AST, U/L872417 (1176–4934)384586 (1759–8580)491771 (1060–3141) <0.0001 Perfusate ALT, U/L881436 (649–2640)382518 (1231–5080)50897 (466–2222) <0.0001 Abbreviations: EAD, early allograft dysfunction; NMP, normothermic machine perfusion.Bold values show statistically significant. Pump variables associated with EAD. Abbreviations: EAD, early allograft dysfunction; NMP, normothermic machine perfusion. Bold values show statistically significant. There was a significant correlation between pump ALT and recipient peak ALT within 7 days (r = 0.60, p < 0.001), with a 1% increase in pump ALT level corresponding to a 0.52% increase in recipient peak ALT level. Similarly, pump AST level was significantly correlated with recipient peak AST level within 7 days (r = 0.59, p < 0.001), with a 1% increase in pump AST corresponding to a 0.55% increase in recipient peak AST (Figure 2). Correlation plots depicting relationships between perfusate and peak recipient transaminases. (A) depicts a positive correlation between pump ALT and peak ALT in the recipient within 7 days, for every 1% increase in pump ALT, peak ALT within 7 days increased by 0.52%. (B) depicts a positive correlation between pump AST and peak AST in the recipient within 7 days, for every 1% increase in pump AST, peak AST within 7 days increased by 0.55%. Recipient Outcomes Associated With EAD One graft was lost to primary non function in this study (0.9%). The recipients in the EAD group had longer hospital stay (10 vs. 8 days, p = 0.03) and higher biopsy proven acute cellular rejection rates (19% vs. 5%, p = 0.02). There were no differences in the degree of renal dysfunction, incidence of infections, incidence of clinically treated acute cellular rejection, degree of rejection, readmission rate, incidence of biliary complications, and need for ERCP between the EAD and non‐EAD groups (Table 3). TABLE 3Outcomes associated with EAD.Total (N = 107)EAD (N = 43)No EAD (N = 64) p‐value N Median (IQR) or n (%) N Median (IQR) or n (%) N Median (IQR) or n (%)Peak ALT within 7 days107536 (267–1085)431081 (714–1576)64351 (164 ‐ 547)NAPeak AST within 7 days1071539 (705–2423)432889 (2279–4379)64932 (494–1410)NABilirubin w/in 7 days1061.7 (1.0–3.5)422.1 (1.0–5.7)641.4 (1.0–2.5)NAINR on POD 71061.1 (1.0–1.2)421.1 (1.0–1.3)641.1 (1.0–1.2)NAPNF1071 (0.9)431 (2.3)640 (0)0.4019AKI w/in 14 days, yes10724 (22.4)439 (20.9)6415 (23.4)0.8169Stage 1 AKI2414 (58.3)95 (55.6)159 (60.0)1.0000Stage 2 AKI4 (16.7)2 (22.2)2 (13.3)Stage 3 AKI6 (25.0)2 (22.2)4 (26.7)Peak Creatinine within 14 days1071.66 (1.06–2.70)431.73 (0.95–2.72)641.65 (1.15–2.69)0.8713New need for HD/ CRRT within 14 days1076 (5.6)433 (7.0)643 (4.7)0.6825Index admission, length of stay, daysICU stay1072.33 (1.88–3.29)432.83 (1.96–3.33)642.19 (1.81–3.19)0.1395hospital stay1079.00 (7.00–12.00)4310.00 (7.00–18.00)648.00 (7.00–11.00) 0.0279 3‐month outcomeseGFR10763 (47–89)4363 (47–89)6465 (46–88)0.8477New need for HD (3 months)1075 (4.7)433 (7.0)642 (3.1)0.3891Bacterial infections1079 (8.4)436 (14.0)643 (4.7)0.1525Fungal infections1072 (1.9)432 (4.7)640 (0)0.15926‐month outcomesAcute cellular rejection (clinically treated)10722 (20.6)439 (20.9)6413 (20.3)1.0000Biopsy proven ACR10611 (10.4)428 (19.0)643 (4.7) 0.0240 Biopsy RAI115 (4–7)85 (4–7)35 (3 ‐ 7)NA12‐month outcomesReadmission8754 (62.1)3626 (72.2)5128 (54.9)0.1202Need for kidney Transplant872 (2.3)361 (2.8)511 (2.0)1.0000Biliary complicationsFollow‐up time, days107378 (263–556)43345 (263–498)64406 (261–597)0.1982ERCP (yes/no)10626 (24.5)4213 (31.0)6413 (20.3)0.2519Number of ERCP procedures1‐22813 (46.4)145 (35.7)148 (57.1)0.60483‐411 (39.3)7 (50.0)4 (28.6)5+4 (14.3)2 (14.3)2 (14.3)Anastomotic stricture, yes10624 (22.6)4212 (28.6)6412 (18.8)0.2470Anastomotic stricture, gradeMild2410 (41.7)125 (41.7)125 (41.7)1.0000Moderate6 (25.0)3 (25.0)3 (25.0)Severe8 (33.3)4 (33.3)4 (33.3)Bile leak, yes1069 (8.5)425 (11.9)644 (6.3)0.4780Ischemic cholangiopathy, yes1067 (6.6)425 (11.9)642 (3.1)0.1107Ischemic cholangiopathy gradeB71 (14.3)50 (NA)21 (50.0)NAC1 (14.3)1 (20.0)0 (NA)D4 (57.1)3 (60.0)1 (50.0)E1 (14.3)1 (20.0)0 (NA)Abbreviations: ACR, acute cellular rejection; AKI, acute kidney injury; CRRT, continuous renal replacement therapy; EAD, early allograft dysfunction; eGFR, estimated glomerular filtration rate; ERCP, Endoscopic retrograde cholangiopancreatography; HD, hemodialysis; ICU, intensive care unit; PNF, primary non function; RAI, rejection activity index.Bold values show statistically significant. Outcomes associated with EAD. Abbreviations: ACR, acute cellular rejection; AKI, acute kidney injury; CRRT, continuous renal replacement therapy; EAD, early allograft dysfunction; eGFR, estimated glomerular filtration rate; ERCP, Endoscopic retrograde cholangiopancreatography; HD, hemodialysis; ICU, intensive care unit; PNF, primary non function; RAI, rejection activity index. Bold values show statistically significant. Descriptive Analysis of Pump Variables Among Four Liver Grafts Lost at 6 Months and Seven Cases of Ischemic Cholangiopathy Nineteen liver grafts demonstrated lactate rebound (rise to >2 mmol/L within 6 h after initial clearance), 69 experienced impaired lactate clearance (<80% decrease by hour 1), 23 had impaired glucose metabolism (<10% decrease in glucose concentration by hour 3), and 41 required >40 mEq of sodium bicarbonate (Table 2). In a single case of PNF, the only notable finding was that the liver required 85 mEq of bicarbonate during NMP. Among the 10 grafts with pump ALT >6000 U/L (n = 10, 90% EAD), none were lost within six months. Of the 17 grafts with AST >6000 U/L (88% EAD), only one failed (Figure 3). A Venn diagram shows the overlap of EAD, 1‐year graft failure, and perfusion threshold, which includes any of the following NMP abnormalities: bicarbonate rise >40 mEq, lactate >2 mmol/L at 6 h, or perfusate ALT >6000 U/L (Figure 3). Missing values were retained in the denominator for percentage calculations (N = 107). Venn Diagram of NMP Perfusion Thresholds, EAD, and 1‑Year Graft Failure. Venn diagram illustrates the overlap between EAD, Perfusion Threshold abnormalities, and 1‑year graft failure. The perfusion threshold includes any of the following NMP abnormalities: bicarbonate rise >40 mEq, lactate >2 mmol/L at 6 h, or perfusate ALT >6000 U/L. Percentages were calculated using a total cohort denominator of N = 107, with missing values retained in the denominator. Discussion Early graft dysfunction scores, such as L‐GrAFT and MEAF scores did not predict short‐term outcomes in deceased donor liver transplants performed after ex‐situ NMP. EAD was significantly associated with inferior 6‐month graft survival (89% vs 100%) in liver grafts that underwent NMP. Higher donor BMI, longer fWIT and anastomosis time, presence of post‐re‐perfusion syndrome, and pump parameters, such as impaired glucose metabolism, delayed lactate clearance, increased use of sodium bicarbonate, and higher perfusate ALT and AST, were associated with EAD. Donor factors traditionally considered to be risk factors for poor outcomes, such as donor age, DCD status, US DRI, and CIT, were not significantly associated with EAD in this dataset. Liver grafts that did not satisfy more conventional viability testing criteria (increased bicarbonate use, increased lactate after reaching nadir, and ALT >6000 U/L) were used with no PNF and excellent short term graft outcomes. NMP has enabled an increase in liver transplant volumes at our center and nationally, particularly from DCD donors [1] At our center, before the use of NMP, strict criteria were enforced when accepting DCD liver grafts. These criteria were primarily based on donor age and anticipated CIT. With NMP, we were able to expand our acceptance criteria, by using liver grafts from older DCD donors, high BMI donors, using local recovery, and “testing” liver quality using NMP prior to implantation. These liver grafts were used if they met the viability criteria. Our evolution in practice patterns can be seen with a significant increase in the US‐DRI scores of DCD liver grafts from 1.82–1.90 in our previous study [20], to 2.43 in our current study. As we have pushed the boundaries, we have become extremely cognizant of patient safety, PNF, and graft loss. There is no national or consensus data currently guiding the use of higher‐risk organs. To help with the development of such guidelines, we meticulously collected and analyzed pump data and performed critical analyses of early graft losses. Fortunately, we did not observe a major uptick in PNF/graft loss, we have reduced time‐to‐transplant, and we are transplanting more recipients than in the pre‐NMP era. The L‐GrAFT and MEAF scores were developed and validated in cohorts managed with conventional (SCS) and relied on the trajectory or cumulative presence (AUC) of early postoperative markers of hepatocellular injury and synthetic function, including transaminases, bilirubin, and INR. With NMP, because IRI occurs more on the pump than in the recipient, the injury profile of the liver graft is fundamentally altered, and the sensitivity of trajectory‐based scores is reduced. In contrast, EAD criteria are fulfilled when fixed, high biochemical thresholds for AST, bilirubin, or INR are exceeded; thus, EAD preferentially identifies grafts that sustain a major early insult, despite NMP. This study had several important limitations. First, its retrospective design and modest sample size restricted the precision of the estimates and limited the strength of the conclusions. Although the cohort included 43 recipients with EAD and 64 without EAD, graft failure and ischemic cholangiopathy were uncommon, with only nine graft failures overall and just one occurring within the first 90 days. Because L‐GrAFT and MEAF scores are designed to predict early graft failure within 90 days posttransplant, the limited number of early events prevented meaningful assessment of their discriminative performance in our population. This will likely be problematic in future studies, as NMP is predicted to significantly reduce the overall incidence of PNF. Similarly, the small number of clinically significant endpoints, including primary non‐function (one case) and liver discards, precluded multivariable modeling and reduced the ability to explore the independent predictors of EAD or graft loss. It is also important to note that we almost always perfused DCD liver grafts under 6 h from cross‐clamp, and within this time frame, CIT did not seem to be a risk factor for EAD. This finding should be interpreted with caution, as emerging data suggest that keeping the time from cross‐clamp to start of perfusion short can significantly reduce the incidence of EAD [21]. Our study was not powered to predict either PNF (one case) or liver discard (5%) as these events occurred infrequently. The absence of bile chemistry could be considered a limitation; however, we did not consider either bile volume or bile chemistry as parameters for liver graft viability. Although biliary bicarbonate, pH, LDH, and glucose have been shown to be markers of bile duct injury on histopathological examination [22], their utility in predicting clinically relevant outcomes remains unanswered [23, 24]. The major strength of our study is that we present outcomes from liver grafts that fall outside the currently accepted viability criteria (perfusate ALT >6000 U/L, rising lactate after initial clearance, and increased use of bicarbonate). Our study also reiterates the importance of lactate metabolism, glucose metabolism, bicarbonate usage, and perfusate transaminase levels as markers of liver dysfunction. We believe the lower MELD‐Na in the EAD group plausibly reflects a selection bias inherent to our practice: when perfusion parameters suggested a graft was more likely to develop EAD, that “marginal” liver was preferentially allocated to a lower‐acuity recipient judged better able to tolerate a rough early recovery. Because of this, our conclusions—particularly the safe utilization of grafts that exceed conventional viability thresholds—may not be directly generalizable to high‐acuity recipients (e.g., MELD‐Na in the 30s, acute‐on‐chronic liver failure, or critically ill/ICU‐bound candidates), in whom diminished physiologic reserve could amplify the consequences of EAD. Due to the small sample size, DCD grafts with NRP followed by ex‐situ NMP were excluded (n = 4), although two patients met the EAD criteria posttransplant. Mechanistically, NRP restores in‐situ oxygenated circulation and can mitigate the IRI that occurs from agonal time. In our data, longer agonal time was significantly associated with EAD in DCD livers. It is therefore biologically plausible that NRP, alone or as a complement to NMP, could reduce EAD‐associated 6‐month graft loss. However, our study was neither designed nor powered to test this hypothesis. Lactate metabolism, an oxygen dependent process, occurs in zone 1 of the hepatic lobules, which are the most resistant to hypoxia. Lactate clearance is a key marker of viability, as failure to do so can be suggestive of severe ischemic injury [25, 26, 27]. Existing criteria for lactate clearance, set as drop in lactate to <2.8 or <2.5 mmol/L within 2 h, support lactate metabolism as the primary criteria for viability [28, 29]. Viability criteria also suggest lactate should not rise after initial clearance. During our early experience, we discarded two liver grafts due to rise in lactate to >2 mmol/L within 6 h after initial lactate clearance to <2 mmol/L. In retrospect, we would have used these two liver grafts. With gained experience, we managed to use 19 liver grafts with a steady rise in lactate levels after initial clearance. This may be related to dependent ischemia from impaired venous outflow due to positioning in the plastic cradle of the NMP [30]. In our experience, we observed a drop in lactate / reduction of the rate of lactate rise after lifting the liver and massaging it. The rate of lactate clearance has also been used as a predictor of liver injury [31]. In our study, temporally impaired lactate clearance was more common in EAD liver grafts. Glycolysis and gluconeogenesis, which are processes of glucose metabolism, are also markers of viability [28, 29]. Glycolysis, a metabolic process that breaks down glucose to pyruvate and generates ATP during aerobic metabolism, predominantly occurs in zone 3 [27, 32]. The slow rate of drop in glucose concentration can be considered a marker of moderate liver injury. In our study, an increased incidence of EAD was observed in liver grafts with less than 10% decrease in glucose concentration by hour 3. Gluconeogenesis, a metabolic process that synthesizes glucose from noncarbohydrate sources, predominantly occurs in zone 1 [27, 32]. Impaired gluconeogenesis should be considered as a marker of severe ischemic injury. In our study, we did not encounter liver grafts that failed to synthesize glucose using amino acids in dextrose (Clinimix, Baxter, Deerfield, IL, USA). This could be due to a selection bias in discarding liver grafts with severe ischemic injury prior to NMP. The amount of bicarbonate required to maintain a perfusate pH >7.2 or ≥7.3 is another viability marker. The ability to maintain pH>7.2 without needing >30 mmol of bicarbonate supplementation, has been proposed as a viability criterion [28, 29]. Hepatic urea synthesis can occur across zones 1, 2 and in the first part of zone 3. Glutamine is hydrolyzed to glutamate and ammonia by glutaminase, and its activity is pH‐dependent and predominantly limited to zone 1. Urea is then produced form ammonia through the urea cycle [27, 33]. Disruption of these processes can lead to the dysregulation of pH and acidemia. It is generally thought that increased doses of bicarbonate needed to correct acidemia can signify moderate to severe ischemia‐reperfusion injury. The optimal pH and the amount of bicarbonate required to maintain the pH as markers of liver function, need to be determined. In our study, 41 liver grafts that required >40 mEq of bicarbonate were used for transplantation, and only one PNF was observed in the liver graft that required 85 mEq of bicarbonate. Higher bicarbonate use was significantly associated with EAD. Perfusate transaminase levels, especially ALT levels of <6000 U/L at 2 h, are also commonly used as viability criteria [7, 34]. Ten grafts with ALT >6000 U/L were used, and none developed PNF or graft loss at 6 months. Transaminase levels are influenced by donor age, graft weight, ischemia time, and steatosis. A major limitation of using perfusate transaminases is that these levels are not usually normalized to liver graft weight and perfusate volume [7]. In conclusion, in the NMP cohort, the “NMP‐EAD” phenotype‐ the perfusion‐derived constellation of impaired glucose metabolism, delayed lactate clearance, increased bicarbonate demand, and elevated perfusate transaminases that accompanies posttransplant EAD‐ appears to denote a biologically distinct, high‐risk subgroup with reduced short‐term graft survival, whereas more granular trajectory‐based scores (L‐GrAFT and MEAF) did not discriminate outcomes to the same extent. These parameters point toward a dynamic perfusion‐based predictive framework that is conceptually distinct from the SCS era scores, and warrants prospective, multicenter validation before clinical adoption. Although transcriptomic signatures during NMP have shown promise in characterizing IRI and predicting early graft function, they remain expensive, require bioinformatics support and biobanking infrastructure, and are therefore not yet feasible for routine clinical use [35]. Similarly, flavin mononucleotide (FMN) quantification has recently been validated as a marker of mitochondrial injury with the ability to predict graft outcomes after NMP; however, its application is currently confined to a small number of centers and has not been widely adopted as a routine viability assessment tool yet [8]. Data from multicenter studies are necessary to validate early graft dysfunction scores and to further refine the current viability criteria to enable maximal graft utilization with minimal patient harm. Liver grafts with increased bicarbonate use to maintain pH >7.2, rise in lactate to >2 mmol/L within 6 h after initial lactate clearance, and perfusate ALT >6000 U/L can be safely used in carefully selected donors and in the context of the entire spectrum of viability markers. Neither lactate clearance nor transaminase thresholds alone are adequate for viability assessment, and utilization of singular thresholds will lead to discard of transplantable livers. Dynamic and composite evaluation of multiple parameters including lactate clearance, bicarbonate demand, and glucose utilization has allowed our center to safely maximize transplant volume without marked increases in primary nonfunction, graft loss, or patient death. Author Contributions Conception or design of work: Kejal Shah, William K Washburn, Austin D Schenk, Musab Alebrahim, and Sai Rithin Punjala. Acquisition, analysis, or interpretation of data for the work: Kejal Shah, April J Logan, Leonid Gorelik, Annelise Nolan, Manoj Iyer, Lauren Von Stein, Wei Chen, Ayato Obana, Ashley Limkemann, Navdeep Singh, Sylvester Black, and Sai Rithin Punjala. Drafting the work: Kejal Shah, April J Logan, Austin D Schenk, Musab Alebrahim, and Sai Rithin Punjala. Critical revision of the manuscript for important intellectual content: Leonid Gorelik, Annelise Nolan, Manoj Iyer, Lauren Von Stein, Wei Chen, Ayato Obana, Ashley Limkemann, Navdeep Singh, Sylvester Black, and William K Washburn. All authors approved the final version to be published and agreed to be accountable for all aspects of the work . Funding The authors have nothing to report. Ethics Statement This study was approved by the Institutional Review Board of The Ohio State University (Protocol #2023H0246). The requirement for individual informed consent was waived owing to the retrospective nature of the study. All research was conducted in accordance with the ethical principles of the Declaration of Helsinki and the Declaration of Istanbul. No organs or tissue were procured from prisoners or from executed persons. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Supporting Information: ctr70632‐sup‐0001‐SuppMat.docx